GitLab AI Accountability Report AI Allows Companies to Write Code Faster than they Can Review It

Related Vendors

dataTec AG

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

SIMTOS (Seoul International Machine Tool Show)

AI coding tools significantly accelerate software development. However, a GitLab study notes that many companies lack the necessary governance, traceability, and clear lines of responsibility.

AI-powered software development has evolved from an experiment to standard practice in many companies. According to GitLab’s AI Accountability Report, conducted by The Harris Poll, 91 percent of the organizations surveyed are already actively using two or more AI coding tools. Fifty-four percent use three or more such tools. The study is based on data from 1,528 DevSecOps professionals across six countries.

Gallery

Gallery with 10 images

The benefits are clearly evident to many respondents. Sixty percent say that the return on investment for their AI coding tools has exceeded expectations. Seventy-eight percent report that developers have been writing and committing code faster since the introduction of AI. Seventy-three percent also see an improvement in the quality of the code that reaches production.

At the same time, the bottleneck in the software delivery process is shifting. According to the report, respondents now spend only 16 percent of their working time writing new code. 85 percent agree with the statement that AI has shifted the bottleneck from writing code to testing and validating it. So far, respondents see particularly few efficiency gains in areas such as compliance, security scanning, deployment, and incident response.

This creates a productivity paradox: While 79 percent of participants see higher individual productivity thanks to AI, they do not see an equally significant improvement across the entire software delivery lifecycle. GitLab therefore no longer defines the next level of maturity solely in terms of faster code generation, but rather in terms of the ability to reliably classify, review, and take responsibility for AI-generated code.

Governance is Becoming A Key Priority

GitLab defines AI accountability as the organizational and technical ability to answer three questions about every AI-generated code component: Where did the code come from, what was it intended for, and who is responsible for it once it goes into production? This is precisely where many companies fall short. 43 percent of respondents say they cannot reliably distinguish AI-generated code from code written by humans.

Additional structural hurdles complicate monitoring. Forty percent cite fragmented toolchains as a problem, while 39 percent cite systems that do not track the origin and purpose of code. Only 28 percent report that their tools are fully integrated across the software development lifecycle and utilize shared data and workflows.

This gap is particularly evident in incidents involving production. While 87 percent of respondents are confident that they can determine within 24 hours whether AI-generated code was involved in a production problem, However, among the organizations that actually experienced an incident last year, 34 percent were unable to provide a definitive answer to this question.

Gallery

Gallery with 10 images

For companies, governance thus becomes a matter of investment. 92 percent report challenges in dealing with AI-generated code, and 80 percent say their organization adopted AI tools faster than it could establish appropriate guidelines. 82 percent see the risk of a new form of technical debt for which their company is not yet sufficiently prepared. At the same time, 91 percent of respondents plan to invest in tools or processes for AI code governance, traceability, or maintainability over the next twelve months.(sg)